An impulse response captures how a speaker, cabinet, microphone, position, and room shape a sound. Amp Rig is a useful teaching example because its cabinet stage follows the same three-step process.
Step one: the sweep
Put a speaker cabinet in a room, put a microphone in front of it, and play a signal through it that contains every frequency you care about. The usual choice is a sine sweep that glides from around 20 Hz to 20 kHz over several seconds. You record what the microphone hears.
What the microphone hears is not the sweep. It is the sweep plus everything the cabinet, the speaker cone, the microphone, the mic position, and the room did to it. The cone rolls off the top end hard. The cabinet resonates somewhere in the low mids. The room adds early reflections. All of it is baked into that recording, indivisibly, which turns out to be the useful part.
Step two: deconvolution
You now have the input (the sweep you sent) and the output (what came back). The remaining task is to recover the transformation the system applied. For a sine sweep, convolve the recording with a time-reversed, amplitude-corrected copy of the sweep. The excitation collapses into a single point in time, leaving the system's response to an ideal impulse.
That result is the impulse response. It is a short audio file, typically tens to a few hundred milliseconds. Played alone, it sounds like a brief click with the tonal imprint of the captured system.
The sweep method is used instead of literally firing a starter pistol at the cabinet because a sweep puts far more total energy into the system while staying inside its linear range, which is the difference between an IR with usable signal-to-noise and one that is mostly hiss.
Step three: convolution
Playing guitar through that cabinet is now a matter of convolving your signal with the IR. Every sample of your input scatters a scaled copy of the impulse response across the output, and the sum is your guitar as that microphone would have heard it in that room.
Done naively, this is expensive because every output sample costs one multiply per IR sample. Real-time convolution uses partitioned FFT blocks instead, splitting the IR into chunks so early chunks are processed at low latency and later chunks are handled in larger, cheaper blocks. Amp Rig uses this approach in the current browser module.
Why a 4x12 flatters a bad take
The cabinet's band limits matter when you are recording.
A guitar cabinet is a strong band-limiting filter. A 4x12 rolls off sharply above roughly 5 kHz and has very little below 80 Hz or so. Much of the fizzy top end from distortion lands in the region the cabinet removes. Pick noise, string squeak, and the harsh edge of a slightly flat bend live there too.
So the cabinet is not a flattering EQ curve added to good playing. It is a band limit that removes most of the frequency range where inaccuracy is audible. Play a sloppy take through a raw distortion stage and you hear every fault. Put a 4x12 after it and a good deal of that evidence is simply gone.
This is worth knowing in both directions. It means a distortion sim judged without a cabinet is being judged unfairly. It also means that if your take sounds great through a heavy cab and thin through anything else, the cab was doing more work than you were.
Bring your own
Amp Rig's current web module can load your own .wav impulses locally, and the file stays on your device. Native plugin and standalone formats remain planned, so the browser is the supported path today.
